Add aglab2 lz4t

Makefile

@@ -315,7 +315,7 @@ export LANG := C
 #==============================================================================#
 
 # N64 tools
-MIO0TOOL              := $(TOOLS_DIR)/sm64tools/mio0
+MIO0TOOL              := $(TOOLS_DIR)/lz4tpack
 N64CKSUM              := $(TOOLS_DIR)/sm64tools/n64cksum
 N64GRAPHICS           := $(TOOLS_DIR)/sm64tools/n64graphics
 N64GRAPHICS_CI        := $(TOOLS_DIR)/sm64tools/n64graphics_ci

asm/decompress.s

@@ -1,104 +1,216 @@
-// assembler directives
-.set noat      // allow manual use of $at
-.set noreorder // don't insert nops after branches
+##############################################
+#  LZ4T - Fast decompressor in assembly
+#  Written by aglab2 inspired by Rasky LZ4 and devwizard YAZ0
+##############################################
 
-#include "macros.inc"
+# NOTE: to optimize for speed, this decompressor can write up to 8 bytes
+# after the end of the output buffer. The outut buffer must have been sized
+# accordingly to accomodate for this.
 
+#define MINMATCH    4
 
-.section .text, "ax"
+#define inbuf       $s0
+#define nibbles     $s1
+#define outbuf      $s2
+#define match_combo_mask $s3
+#define len         $s4
+#define match_lim   $s5
+#define match_min   $s6
+#define v0_st       $s7
 
-// This file is handwritten.
+#define dma_ctx     $s8
+#define dma_ptr     $v0
 
-glabel decompress
-#if !defined(VERSION_JP) && !defined(VERSION_US)
-    lw    $a3, 8($a0)
-    lw    $t9, 0xc($a0)
-    lw    $t8, 4($a0)
-    add   $a3, $a3, $a0
-    add   $t9, $t9, $a0
-    or    $a2, $zero, $zero
-    addi  $a0, $a0, 0x10
-    add   $t8, $t8, $a1
-.L8026ED80:
-    bnezl $a2, .L8026ED98
-     slt   $t1, $t0, $zero
-    lw    $t0, ($a0)
-    li    $a2, 32
-    addi  $a0, $a0, 4
-    slt   $t1, $t0, $zero
-.L8026ED98:
-    beql  $t1, $zero, .L8026EDB8
-     lhu   $t2, ($a3)
-    lb    $t2, ($t9)
-    addi  $t9, $t9, 1
-    addi  $a1, $a1, 1
-    b     .L8026EDE4
-     sb    $t2, -1($a1)
-    lhu   $t2, ($a3)
-.L8026EDB8:
-    addi  $a3, $a3, 2
-    srl   $t3, $t2, 0xc
-    andi  $t2, $t2, 0xfff
-    sub   $t1, $a1, $t2
-    addi  $t3, $t3, 3
-.L8026EDCC:
-    lb    $t2, -1($t1)
-    addi  $t3, $t3, -1
-    addi  $t1, $t1, 1
-    addi  $a1, $a1, 1
-    bnez  $t3, .L8026EDCC
-     sb    $t2, -1($a1)
-.L8026EDE4:
-    sll   $t0, $t0, 1
-    bne   $a1, $t8, .L8026ED80
-     addi  $a2, $a2, -1
-    jr    $ra
+#define shift       $t9
+#define len_add     $t8
+#define match_len   $t7
+#define match_off   $t6
+#define match_combo $t5
+#define off_nibble  $t4
+
+    .section .text.lz4t_unpack_fast
+	.p2align 5
+    .globl lz4t_unpack_fast 
+    .func lz4t_unpack_fast
+    .set at
+    .set noreorder
+
+lz4t_unpack_fast:
+    addiu $sp, $sp, -0x40
+    sw $ra, 0x14($sp)
+    sw $s0, 0x18($sp)
+    sw $s1, 0x1c($sp)
+    sw $s2, 0x20($sp)
+    sw $s3, 0x24($sp)
+    sw $s4, 0x28($sp)
+    sw $s5, 0x2C($sp)
+    sw $s6, 0x30($sp)
+    sw $s7, 0x34($sp)
+    sw $s8, 0x38($sp)
+
+    move $s0, $a0
+    lw $s1, 12($a0)
+    move $s2, $a1
+    move dma_ctx, $a2
+    lbu match_combo_mask, 8($a0)
+    sll match_combo_mask, 28
+    lbu match_min, 9($a0)
+
+    move dma_ptr, $a0
+    addiu $s0, 16
+
+.Lloop:
+    sub $t0, inbuf, dma_ptr                     # check if we need to wait for dma
+    bgezal $t0, dma_async_ctx_read                    # if inbuf >= dma_ptr, wait for dma
+     move $a0, dma_ctx
+
+    bnez nibbles, .Lprocess_nibbles
+     li match_lim, 7
+     
+.Lload_nibbles:
+    lwl nibbles, 0(inbuf)
+    lwr nibbles, 3(inbuf)
+    beqz nibbles, .Lend
+    add inbuf, 4
+
+.Lprocess_nibbles:
+    bgez nibbles, .Lmatches
+     srl len, nibbles, 28
+
+.Lliterals:
+    andi len, 7
+    beqz len, .Llarge_literals
      nop
-#else
-    lw    $t8, 4($a0)
-    lw    $a3, 8($a0)
-    lw    $t9, 0xc($a0)
-    move  $a2, $zero
-    add   $t8, $t8, $a1
-    add   $a3, $a3, $a0
-    add   $t9, $t9, $a0
-    addi  $a0, $a0, 0x10
-.L8027EF50:
-    bnez  $a2, .L8027EF64
+
+.Lsmall_literal:
+    ldl $t0, 0(inbuf)
+    ldr $t0, 7(inbuf)
+    add inbuf, len
+    sdl $t0, 0(outbuf)
+    sdr $t0, 7(outbuf)
+    sll nibbles, 4
+    beq len, match_lim, .Lloop
+    add outbuf, len
+
+.Lmatches_ex:
+    sub $t0, inbuf, dma_ptr                     # check if we need to wait for dma
+    bgezal $t0, dma_async_ctx_read                    # if inbuf >= dma_ptr, wait for dma
+     move $a0, dma_ctx
+
+    bnez nibbles, .Lprocess_ex_match_nibble
+     li match_lim, 15
+     
+.Lload_nibbles2:
+    lwl nibbles, 0(inbuf)
+    lwr nibbles, 3(inbuf)
+    beqz nibbles, .Lend
+    add inbuf, 4
+
+.Lprocess_ex_match_nibble:
+    srl len, nibbles, 28
+
+.Lmatches:
+    lwl match_combo, 0(inbuf)
+    lwr match_combo, 3(inbuf)
+    addiu inbuf, 2
+    srl match_off, match_combo, 16
+
+    beqz match_combo_mask, .Lfull_offset
+    sll nibbles, 4
+    srl nibbles, 4
+    and off_nibble, match_combo, match_combo_mask
+    or nibbles, off_nibble
+    andi match_off, 0xfff
+.Lfull_offset:
+
+    bne len, match_lim, .Lmatch
+     addu match_len, len, match_min
+
+    # len is sign extended match_combo[8:15]
+    sll match_combo, 16
+    sra len, match_combo, 24
+    add inbuf, 1
+    bltzal len, .Lread_large_amount
+     andi len, 0x7f
+
+    add match_len, len
+
+.Lmatch:
+    ble match_off, match_len, .Lmatch1_loop     # check if we can do 8-byte copy
+     sub v0_st, outbuf, match_off                 # calculate start of match
+.Lmatch8_loop:                                  # 8-byte copy loop
+    ldl $t0, -1(v0_st)                             # load 8 bytes
+    ldr $t0, 6(v0_st)
+    addiu v0_st, 8
+    sdl $t0, 0(outbuf)                          # store 8 bytes
+    sdr $t0, 7(outbuf)
+    addiu match_len, -8
+    bgtz match_len, .Lmatch8_loop               # check we went past match_len
+     addiu outbuf, 8
+    b .Lloop                                    # jump to main loop
+     addu outbuf, match_len                     # adjust pointer remove extra bytes
+
+.Lmatch1_loop:                                  # 1-byte copy loop
+    lbu $t0, -1(v0_st)                             # load 1 byte
+    addiu v0_st, 1
+    sb $t0, 0(outbuf)                           # store 1 byte
+    addiu match_len, -1
+    bgtz match_len, .Lmatch1_loop               # check we went past match_len
+     addiu outbuf, 1
+    b .Lloop                                    # jump to main loop
      nop
-    lw    $t0, ($a0)
-    li    $a2, 32
-    addi  $a0, $a0, 4
-.L8027EF64:
-    slt   $t1, $t0, $zero
-    beqz  $t1, .L8027EF88
-     nop
-    lb    $t2, ($t9)
-    addi  $t9, $t9, 1
-    sb    $t2, ($a1)
-    addi  $a1, $a1, 1
-    b     .L8027EFBC
-     nop
-.L8027EF88:
-    lhu   $t2, ($a3)
-    addi  $a3, $a3, 2
-    srl   $t3, $t2, 0xc
-    andi  $t2, $t2, 0xfff
-    sub   $t1, $a1, $t2
-    addi  $t3, $t3, 3
-.L8027EFA0:
-    lb    $t2, -1($t1)
-    addi  $t3, $t3, -1
-    addi  $t1, $t1, 1
-    sb    $t2, ($a1)
-    addi  $a1, $a1, 1
-    bnez  $t3, .L8027EFA0
-     nop
-.L8027EFBC:
-    sll   $t0, $t0, 1
-    addi  $a2, $a2, -1
-    bne   $a1, $t8, .L8027EF50
-     nop
-    jr    $ra
-     nop
-#endif
+
+.Llarge_literals:
+    lb len, 0(inbuf)
+    add inbuf, 1
+    bltzal len, .Lread_large_amount
+     andi len, 0x7f
+
+    move v0_st, inbuf                            # store start of literals into v0_st
+    addiu len, 22
+    add inbuf, len                        # advance inbuf to end of literals
+.Lcopy_lit:
+    sub $t0, v0_st, dma_ptr                     # check if all the literals have been DMA'd
+    bgezal $t0, dma_async_ctx_read                       # if not, wait for DMA
+     move $a0, dma_ctx
+    ldl $t0, 0(v0_st)                             # load 8 bytes of literals
+    ldr $t0, 7(v0_st)
+    addiu v0_st, 8
+    sdl $t0, 0(outbuf)                          # store 8 bytes of literals
+    sdr $t0, 7(outbuf)
+    addiu len, -8
+    bgez len, .Lcopy_lit                  # check if we went past the end of literals
+     addiu outbuf, 8
+    addu outbuf, len                      # adjust outbuf to roll back extra copied bytes
+
+    b .Lmatches_ex
+    sll nibbles, 4
+
+.Lend:
+    lw $ra, 0x14($sp)
+    lw $s0, 0x18($sp)
+    lw $s1, 0x1c($sp)
+    lw $s2, 0x20($sp)
+    lw $s3, 0x24($sp)
+    lw $s4, 0x28($sp)
+    lw $s5, 0x2C($sp)
+    lw $s6, 0x30($sp)
+    lw $s7, 0x34($sp)
+    lw $s8, 0x38($sp)
+    jr $ra
+    addiu $sp, $sp, 0x40
+
+.Lread_large_amount:
+    li shift, 7
+.Lread_large_amount_loop:
+    lb $t0, 0(inbuf)
+    add inbuf, 1
+    andi $t1, $t0, 0x7f
+    sllv $t1, $t1, shift
+    or len, $t1
+    bltz $t0, .Lread_large_amount_loop
+    add shift, 7
+    jr $ra
+    nop
+
+.endfunc

sm64.ld

@@ -46,7 +46,8 @@ SECTIONS
     BEGIN_SEG(main, .)
     {
         BUILD_DIR/asm/entry.o(.text);
-        BUILD_DIR/asm/decompress.o(.text);
+        BUILD_DIR/asm/decompress.o(.text*);
+        BUILD_DIR/src/game/dma_async.o(.text);
         BUILD_DIR/src/game*.o(.text);
         AUDIO_DIR*.o(.text);
         ULTRA_BUILD_DIR/libgultra_rom.a:*.o(.text);
@@ -55,6 +56,7 @@ SECTIONS
 
         BUILD_DIR/asm/entry.o(.data*);
         BUILD_DIR/asm/decompress.o(.data*);
+        BUILD_DIR/src/game/dma_async.o(.data);
         BUILD_DIR/src/game*.o(.data*);
         AUDIO_DIR*.o(.data*);
         ULTRA_BUILD_DIR/libgultra_rom.a:*.o(.data*);
@@ -63,6 +65,7 @@ SECTIONS
 
         BUILD_DIR/asm/entry.o(.rodata*);
         BUILD_DIR/asm/decompress.o(.rodata*);
+        BUILD_DIR/src/game/dma_async.o(.rodata);
         BUILD_DIR/src/game*.o(.rodata*);
         AUDIO_DIR*.o(.rodata*);
         ULTRA_BUILD_DIR/libgultra_rom.a:*.o(.rodata*);

src/buffers/buffers.c

@@ -2,8 +2,6 @@
 
 #include "buffers.h"
 
-ALIGNED8 u8 gDecompressionHeap[0xD000];
-
 #if defined(VERSION_EU)
 ALIGNED16 u8 gAudioHeap[DOUBLE_SIZE_ON_64_BIT(0x31200) - 0x3800];
 #elif defined(VERSION_SH)

src/buffers/buffers.h

@@ -7,8 +7,6 @@
 #include "engine/game_init.h"
 #include "config.h"
 
-extern u8 gDecompressionHeap[];
-
 extern u8 gAudioHeap[];
 
 extern u8 gAudioSPTaskYieldBuffer[];

src/engine/level_script.c

@@ -297,7 +297,7 @@ static void level_cmd_load_mario_head(void) {
 }
 
 static void level_cmd_load_mio0_texture(void) {
-    load_segment_decompress_heap(CMD_GET(s16, 2), CMD_GET(void *, 4), CMD_GET(void *, 8));
+    load_segment_decompress(CMD_GET(s16, 2), CMD_GET(void *, 4), CMD_GET(void *, 8));
     sCurrentCmd = CMD_NEXT;
 }
 

src/game/decompress.h

@@ -1,6 +1,7 @@
 #ifndef DECOMPRESS_H
 #define DECOMPRESS_H
 
-void decompress(void *mio0, void *dest);
+void lz4t_unpack_fast(const uint8_t* restrict inbuf, uint8_t* restrict dst, DMAAsyncCtx *ctx);
+#define DMA_ASYNC_HEADER_SIZE 16
 
 #endif // DECOMPRESS_H

src/game/dma_async.c

@@ -0,0 +1,42 @@
+#include "dma_async.h"
+#include "game/main.h"
+
+#define ALIGN16(val) (((val) + 0xF) & ~0xF)
+
+void dma_async_ctx_init(DMAAsyncCtx* ctx, u8 *dest, u8 *srcStart, u8 *srcEnd) {
+    u32 size = ALIGN16(srcEnd - srcStart);
+    osInvalDCache(dest, size);
+
+    u32 copySize = (size >= 0x1000) ? 0x1000 : size;
+
+    osPiStartDma(&gDmaIoMesg, OS_MESG_PRI_NORMAL, OS_READ, (uintptr_t) srcStart, dest, copySize, &gDmaMesgQueue);
+
+    dest += copySize;
+    srcStart += copySize;
+    size -= copySize;
+
+    ctx->srcStart = srcStart;
+    ctx->dest = dest;
+    ctx->size = size;
+}
+
+void* dma_async_ctx_read(DMAAsyncCtx* ctx) {
+    // wait for the previous DMA issued
+    osRecvMesg(&gDmaMesgQueue, &gMainReceivedMesg, OS_MESG_BLOCK);
+
+    // start the new DMA transfer
+    u32 copySize = (ctx->size >= 0x1000) ? 0x1000 : ctx->size;
+    if (copySize == 0) {
+        // we are done, return a dummy address that is so gigantic that we will never be called again
+        return (void*) 0x80800000;
+    }
+    osPiStartDma(&gDmaIoMesg, OS_MESG_PRI_NORMAL, OS_READ, (uintptr_t) ctx->srcStart, ctx->dest, copySize, &gDmaMesgQueue);
+
+    const u32 margin = 16;
+    void* ret = ctx->dest - margin;
+    ctx->dest += copySize;
+    ctx->srcStart += copySize;
+    ctx->size -= copySize;
+
+    return ret;
+}

src/game/dma_async.h

@@ -0,0 +1,15 @@
+#pragma once
+
+#include "types.h"
+
+typedef struct {
+    u8* srcStart;
+    u8* dest;
+    u32 size;
+} DMAAsyncCtx;
+
+// Starts to DMA the first block
+void dma_async_ctx_init(DMAAsyncCtx* ctx, u8 *dest, u8 *srcStart, u8 *srcEnd);
+
+// Starts to DMA the next block and waits for the previous block
+void* dma_async_ctx_read(DMAAsyncCtx* ctx);

src/game/libgcc.c

@@ -183,6 +183,35 @@ int __ucmpdi2(unsigned long long a, unsigned long long b) {
     return (a < b) ? 0 : 2;
 }
 
+// Taken from LLVM
+
+typedef union {
+    u64 all;
+    struct {
+        u32 high;
+        u32 low;
+    } s;
+} UdWords;
+
+s64 __lshrdi3(s64 a, s32 b) {
+    const s32 bits_in_word = (s32)(sizeof(s32) * 8);
+    UdWords input;
+    UdWords result;
+    input.all = a;
+    if (b & bits_in_word)  /* bits_in_word <= b < bits_in_dword */ {
+        result.s.high = 0;
+        result.s.low = input.s.high >> (b - bits_in_word);
+    } else  /* 0 <= b < bits_in_word */ {
+        if (b == 0) {
+            return a;
+        }
+
+        result.s.high  = input.s.high >> b;
+        result.s.low = (input.s.high << (bits_in_word - b)) | (input.s.low >> b);
+    }
+    return result.all;
+}
+
 // Compute division and modulo of 64-bit signed and unsigned integers
 
 __asm__("                                   \n\

src/game/memory.c

@@ -6,6 +6,7 @@
 
 #include "buffers/zbuffer.h"
 #include "buffers/buffers.h"
+#include "dma_async.h"
 #include "decompress.h"
 #include "engine/game_init.h"
 #include "main.h"
@@ -342,7 +343,10 @@ void *load_segment_decompress(s32 segment, u8 *srcStart, u8 *srcEnd) {
         dest = main_pool_alloc(*size, MEMORY_POOL_LEFT);
         if (dest != NULL) {
             CN_DEBUG_PRINTF(("start decompress\n"));
-            decompress(compressed, dest);
+            dma_read(compressed, srcStart, srcStart + DMA_ASYNC_HEADER_SIZE);
+            DMAAsyncCtx asyncCtx;
+            dma_async_ctx_init(&asyncCtx, compressed + DMA_ASYNC_HEADER_SIZE, srcStart + DMA_ASYNC_HEADER_SIZE, srcEnd);
+            lz4t_unpack_fast(compressed, dest, &asyncCtx);
             CN_DEBUG_PRINTF(("end decompress\n"));
 
             set_segment_base_addr(segment, dest);
@@ -354,22 +358,6 @@ void *load_segment_decompress(s32 segment, u8 *srcStart, u8 *srcEnd) {
     return dest;
 }
 
-void *load_segment_decompress_heap(u32 segment, u8 *srcStart, u8 *srcEnd) {
-    UNUSED void *dest = NULL;
-    u32 compSize = ALIGN16(srcEnd - srcStart);
-    u8 *compressed = main_pool_alloc(compSize, MEMORY_POOL_RIGHT);
-    UNUSED u32 *pUncSize = (u32 *) (compressed + 4);
-
-    if (compressed != NULL) {
-        dma_read(compressed, srcStart, srcEnd);
-        decompress(compressed, gDecompressionHeap);
-        set_segment_base_addr(segment, gDecompressionHeap);
-        main_pool_free(compressed);
-    } else {
-    }
-    return gDecompressionHeap;
-}
-
 #ifndef LIBDRAGON_IPL3
 void load_engine_code_segment(void) {
     void *startAddr = (void *) _engineSegmentStart;

tools/Makefile

@@ -7,7 +7,7 @@ CC           := gcc
 CXX          := g++
 CFLAGS       := -I . -I sm64tools -Wall -Wextra -Wno-unused-parameter -pedantic -O2 -s
 LDFLAGS      := -lm
-ALL_PROGRAMS := armips textconv patch_elf_32bit aifc_decode aiff_extract_codebook vadpcm_enc tabledesign extract_data_for_mio skyconv
+ALL_PROGRAMS := armips textconv patch_elf_32bit aifc_decode aiff_extract_codebook vadpcm_enc tabledesign extract_data_for_mio skyconv lz4tpack
 LIBAUDIOFILE := audiofile/libaudiofile.a
 
 # Only build armips from tools if it is not found on the system
@@ -19,6 +19,9 @@ endif
 
 default: all
 
+lz4tpack_SOURCES := lz4tpack.c
+lz4tpack_CFLAGS  := -DLZ4T -flto -O3 -fwhole-program
+
 textconv_SOURCES := textconv.c utf8.c hashtable.c
 
 patch_elf_32bit_SOURCES := patch_elf_32bit.c

tools/lz4hc.h

@@ -0,0 +1,413 @@
+/*
+   LZ4 HC - High Compression Mode of LZ4
+   Header File
+   Copyright (C) 2011-2020, Yann Collet.
+   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+   Redistribution and use in source and binary forms, with or without
+   modification, are permitted provided that the following conditions are
+   met:
+
+       * Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+       * Redistributions in binary form must reproduce the above
+   copyright notice, this list of conditions and the following disclaimer
+   in the documentation and/or other materials provided with the
+   distribution.
+
+   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+   You can contact the author at :
+   - LZ4 source repository : https://github.com/lz4/lz4
+   - LZ4 public forum : https://groups.google.com/forum/#!forum/lz4c
+*/
+#ifndef LZ4_HC_H_19834876238432
+#define LZ4_HC_H_19834876238432
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* --- Dependency --- */
+/* note : lz4hc requires lz4.h/lz4.c for compilation */
+#include "lz4.h"   /* stddef, LZ4LIB_API, LZ4_DEPRECATED */
+
+
+/* --- Useful constants --- */
+#define LZ4HC_CLEVEL_MIN         2
+#define LZ4HC_CLEVEL_DEFAULT     9
+#define LZ4HC_CLEVEL_OPT_MIN    10
+#define LZ4HC_CLEVEL_MAX        12
+
+
+/*-************************************
+ *  Block Compression
+ **************************************/
+/*! LZ4_compress_HC() :
+ *  Compress data from `src` into `dst`, using the powerful but slower "HC" algorithm.
+ * `dst` must be already allocated.
+ *  Compression is guaranteed to succeed if `dstCapacity >= LZ4_compressBound(srcSize)` (see "lz4.h")
+ *  Max supported `srcSize` value is LZ4_MAX_INPUT_SIZE (see "lz4.h")
+ * `compressionLevel` : any value between 1 and LZ4HC_CLEVEL_MAX will work.
+ *                      Values > LZ4HC_CLEVEL_MAX behave the same as LZ4HC_CLEVEL_MAX.
+ * @return : the number of bytes written into 'dst'
+ *           or 0 if compression fails.
+ */
+LZ4LIB_API int LZ4_compress_HC (const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel);
+
+
+/* Note :
+ *   Decompression functions are provided within "lz4.h" (BSD license)
+ */
+
+
+/*! LZ4_compress_HC_extStateHC() :
+ *  Same as LZ4_compress_HC(), but using an externally allocated memory segment for `state`.
+ * `state` size is provided by LZ4_sizeofStateHC().
+ *  Memory segment must be aligned on 8-bytes boundaries (which a normal malloc() should do properly).
+ */
+LZ4LIB_API int LZ4_sizeofStateHC(void);
+LZ4LIB_API int LZ4_compress_HC_extStateHC(void* stateHC, const char* src, char* dst, int srcSize, int maxDstSize, int compressionLevel);
+
+
+/*! LZ4_compress_HC_destSize() : v1.9.0+
+ *  Will compress as much data as possible from `src`
+ *  to fit into `targetDstSize` budget.
+ *  Result is provided in 2 parts :
+ * @return : the number of bytes written into 'dst' (necessarily <= targetDstSize)
+ *           or 0 if compression fails.
+ * `srcSizePtr` : on success, *srcSizePtr is updated to indicate how much bytes were read from `src`
+ */
+LZ4LIB_API int LZ4_compress_HC_destSize(void* stateHC,
+                                  const char* src, char* dst,
+                                        int* srcSizePtr, int targetDstSize,
+                                        int compressionLevel);
+
+
+/*-************************************
+ *  Streaming Compression
+ *  Bufferless synchronous API
+ **************************************/
+ typedef union LZ4_streamHC_u LZ4_streamHC_t;   /* incomplete type (defined later) */
+
+/*! LZ4_createStreamHC() and LZ4_freeStreamHC() :
+ *  These functions create and release memory for LZ4 HC streaming state.
+ *  Newly created states are automatically initialized.
+ *  A same state can be used multiple times consecutively,
+ *  starting with LZ4_resetStreamHC_fast() to start a new stream of blocks.
+ */
+LZ4LIB_API LZ4_streamHC_t* LZ4_createStreamHC(void);
+LZ4LIB_API int             LZ4_freeStreamHC (LZ4_streamHC_t* streamHCPtr);
+
+/*
+  These functions compress data in successive blocks of any size,
+  using previous blocks as dictionary, to improve compression ratio.
+  One key assumption is that previous blocks (up to 64 KB) remain read-accessible while compressing next blocks.
+  There is an exception for ring buffers, which can be smaller than 64 KB.
+  Ring-buffer scenario is automatically detected and handled within LZ4_compress_HC_continue().
+
+  Before starting compression, state must be allocated and properly initialized.
+  LZ4_createStreamHC() does both, though compression level is set to LZ4HC_CLEVEL_DEFAULT.
+
+  Selecting the compression level can be done with LZ4_resetStreamHC_fast() (starts a new stream)
+  or LZ4_setCompressionLevel() (anytime, between blocks in the same stream) (experimental).
+  LZ4_resetStreamHC_fast() only works on states which have been properly initialized at least once,
+  which is automatically the case when state is created using LZ4_createStreamHC().
+
+  After reset, a first "fictional block" can be designated as initial dictionary,
+  using LZ4_loadDictHC() (Optional).
+
+  Invoke LZ4_compress_HC_continue() to compress each successive block.
+  The number of blocks is unlimited.
+  Previous input blocks, including initial dictionary when present,
+  must remain accessible and unmodified during compression.
+
+  It's allowed to update compression level anytime between blocks,
+  using LZ4_setCompressionLevel() (experimental).
+
+  'dst' buffer should be sized to handle worst case scenarios
+  (see LZ4_compressBound(), it ensures compression success).
+  In case of failure, the API does not guarantee recovery,
+  so the state _must_ be reset.
+  To ensure compression success
+  whenever `dst` buffer size cannot be made >= LZ4_compressBound(),
+  consider using LZ4_compress_HC_continue_destSize().
+
+  Whenever previous input blocks can't be preserved unmodified in-place during compression of next blocks,
+  it's possible to copy the last blocks into a more stable memory space, using LZ4_saveDictHC().
+  Return value of LZ4_saveDictHC() is the size of dictionary effectively saved into 'safeBuffer' (<= 64 KB)
+
+  After completing a streaming compression,
+  it's possible to start a new stream of blocks, using the same LZ4_streamHC_t state,
+  just by resetting it, using LZ4_resetStreamHC_fast().
+*/
+
+LZ4LIB_API void LZ4_resetStreamHC_fast(LZ4_streamHC_t* streamHCPtr, int compressionLevel);   /* v1.9.0+ */
+LZ4LIB_API int  LZ4_loadDictHC (LZ4_streamHC_t* streamHCPtr, const char* dictionary, int dictSize);
+
+LZ4LIB_API int LZ4_compress_HC_continue (LZ4_streamHC_t* streamHCPtr,
+                                   const char* src, char* dst,
+                                         int srcSize, int maxDstSize);
+
+/*! LZ4_compress_HC_continue_destSize() : v1.9.0+
+ *  Similar to LZ4_compress_HC_continue(),
+ *  but will read as much data as possible from `src`
+ *  to fit into `targetDstSize` budget.
+ *  Result is provided into 2 parts :
+ * @return : the number of bytes written into 'dst' (necessarily <= targetDstSize)
+ *           or 0 if compression fails.
+ * `srcSizePtr` : on success, *srcSizePtr will be updated to indicate how much bytes were read from `src`.
+ *           Note that this function may not consume the entire input.
+ */
+LZ4LIB_API int LZ4_compress_HC_continue_destSize(LZ4_streamHC_t* LZ4_streamHCPtr,
+                                           const char* src, char* dst,
+                                                 int* srcSizePtr, int targetDstSize);
+
+LZ4LIB_API int LZ4_saveDictHC (LZ4_streamHC_t* streamHCPtr, char* safeBuffer, int maxDictSize);
+
+
+
+/*^**********************************************
+ * !!!!!!   STATIC LINKING ONLY   !!!!!!
+ ***********************************************/
+
+/*-******************************************************************
+ * PRIVATE DEFINITIONS :
+ * Do not use these definitions directly.
+ * They are merely exposed to allow static allocation of `LZ4_streamHC_t`.
+ * Declare an `LZ4_streamHC_t` directly, rather than any type below.
+ * Even then, only do so in the context of static linking, as definitions may change between versions.
+ ********************************************************************/
+
+#define LZ4HC_DICTIONARY_LOGSIZE 16
+#define LZ4HC_MAXD (1<<LZ4HC_DICTIONARY_LOGSIZE)
+#define LZ4HC_MAXD_MASK (LZ4HC_MAXD - 1)
+
+#define LZ4HC_HASH_LOG 15
+#define LZ4HC_HASHTABLESIZE (1 << LZ4HC_HASH_LOG)
+#define LZ4HC_HASH_MASK (LZ4HC_HASHTABLESIZE - 1)
+
+
+/* Never ever use these definitions directly !
+ * Declare or allocate an LZ4_streamHC_t instead.
+**/
+typedef struct LZ4HC_CCtx_internal LZ4HC_CCtx_internal;
+struct LZ4HC_CCtx_internal
+{
+    LZ4_u32   hashTable[LZ4HC_HASHTABLESIZE];
+    LZ4_u16   chainTable[LZ4HC_MAXD];
+    const LZ4_byte* end;       /* next block here to continue on current prefix */
+    const LZ4_byte* prefixStart;  /* Indexes relative to this position */
+    const LZ4_byte* dictStart; /* alternate reference for extDict */
+    LZ4_u32   dictLimit;       /* below that point, need extDict */
+    LZ4_u32   lowLimit;        /* below that point, no more dict */
+    LZ4_u32   nextToUpdate;    /* index from which to continue dictionary update */
+    short     compressionLevel;
+    LZ4_i8    favorDecSpeed;   /* favor decompression speed if this flag set,
+                                  otherwise, favor compression ratio */
+    LZ4_i8    dirty;           /* stream has to be fully reset if this flag is set */
+    const LZ4HC_CCtx_internal* dictCtx;
+};
+
+#define LZ4_STREAMHC_MINSIZE  262200  /* static size, for inter-version compatibility */
+union LZ4_streamHC_u {
+    char minStateSize[LZ4_STREAMHC_MINSIZE];
+    LZ4HC_CCtx_internal internal_donotuse;
+}; /* previously typedef'd to LZ4_streamHC_t */
+
+/* LZ4_streamHC_t :
+ * This structure allows static allocation of LZ4 HC streaming state.
+ * This can be used to allocate statically on stack, or as part of a larger structure.
+ *
+ * Such state **must** be initialized using LZ4_initStreamHC() before first use.
+ *
+ * Note that invoking LZ4_initStreamHC() is not required when
+ * the state was created using LZ4_createStreamHC() (which is recommended).
+ * Using the normal builder, a newly created state is automatically initialized.
+ *
+ * Static allocation shall only be used in combination with static linking.
+ */
+
+/* LZ4_initStreamHC() : v1.9.0+
+ * Required before first use of a statically allocated LZ4_streamHC_t.
+ * Before v1.9.0 : use LZ4_resetStreamHC() instead
+ */
+LZ4LIB_API LZ4_streamHC_t* LZ4_initStreamHC(void* buffer, size_t size);
+
+
+/*-************************************
+*  Deprecated Functions
+**************************************/
+/* see lz4.h LZ4_DISABLE_DEPRECATE_WARNINGS to turn off deprecation warnings */
+
+/* deprecated compression functions */
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC               (const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC_limitedOutput (const char* source, char* dest, int inputSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC2              (const char* source, char* dest, int inputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC2_limitedOutput(const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC_withStateHC               (void* state, const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC_limitedOutput_withStateHC (void* state, const char* source, char* dest, int inputSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC2_withStateHC              (void* state, const char* source, char* dest, int inputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC2_limitedOutput_withStateHC(void* state, const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC_continue               (LZ4_streamHC_t* LZ4_streamHCPtr, const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC_limitedOutput_continue (LZ4_streamHC_t* LZ4_streamHCPtr, const char* source, char* dest, int inputSize, int maxOutputSize);
+
+/* Obsolete streaming functions; degraded functionality; do not use!
+ *
+ * In order to perform streaming compression, these functions depended on data
+ * that is no longer tracked in the state. They have been preserved as well as
+ * possible: using them will still produce a correct output. However, use of
+ * LZ4_slideInputBufferHC() will truncate the history of the stream, rather
+ * than preserve a window-sized chunk of history.
+ */
+#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION)
+LZ4_DEPRECATED("use LZ4_createStreamHC() instead") LZ4LIB_API void* LZ4_createHC (const char* inputBuffer);
+LZ4_DEPRECATED("use LZ4_freeStreamHC() instead") LZ4LIB_API   int   LZ4_freeHC (void* LZ4HC_Data);
+#endif
+LZ4_DEPRECATED("use LZ4_saveDictHC() instead") LZ4LIB_API     char* LZ4_slideInputBufferHC (void* LZ4HC_Data);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC2_continue               (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC2_limitedOutput_continue (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_createStreamHC() instead") LZ4LIB_API int   LZ4_sizeofStreamStateHC(void);
+LZ4_DEPRECATED("use LZ4_initStreamHC() instead") LZ4LIB_API  int   LZ4_resetStreamStateHC(void* state, char* inputBuffer);
+
+
+/* LZ4_resetStreamHC() is now replaced by LZ4_initStreamHC().
+ * The intention is to emphasize the difference with LZ4_resetStreamHC_fast(),
+ * which is now the recommended function to start a new stream of blocks,
+ * but cannot be used to initialize a memory segment containing arbitrary garbage data.
+ *
+ * It is recommended to switch to LZ4_initStreamHC().
+ * LZ4_resetStreamHC() will generate deprecation warnings in a future version.
+ */
+LZ4LIB_API void LZ4_resetStreamHC (LZ4_streamHC_t* streamHCPtr, int compressionLevel);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* LZ4_HC_H_19834876238432 */
+
+
+/*-**************************************************
+ * !!!!!     STATIC LINKING ONLY     !!!!!
+ * Following definitions are considered experimental.
+ * They should not be linked from DLL,
+ * as there is no guarantee of API stability yet.
+ * Prototypes will be promoted to "stable" status
+ * after successful usage in real-life scenarios.
+ ***************************************************/
+#ifdef LZ4_HC_STATIC_LINKING_ONLY   /* protection macro */
+#ifndef LZ4_HC_SLO_098092834
+#define LZ4_HC_SLO_098092834
+
+#define LZ4_STATIC_LINKING_ONLY   /* LZ4LIB_STATIC_API */
+#include "lz4.h"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*! LZ4_setCompressionLevel() : v1.8.0+ (experimental)
+ *  It's possible to change compression level
+ *  between successive invocations of LZ4_compress_HC_continue*()
+ *  for dynamic adaptation.
+ */
+LZ4LIB_STATIC_API void LZ4_setCompressionLevel(
+    LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel);
+
+/*! LZ4_favorDecompressionSpeed() : v1.8.2+ (experimental)
+ *  Opt. Parser will favor decompression speed over compression ratio.
+ *  Only applicable to levels >= LZ4HC_CLEVEL_OPT_MIN.
+ */
+LZ4LIB_STATIC_API void LZ4_favorDecompressionSpeed(
+    LZ4_streamHC_t* LZ4_streamHCPtr, int favor);
+
+/*! LZ4_resetStreamHC_fast() : v1.9.0+
+ *  When an LZ4_streamHC_t is known to be in a internally coherent state,
+ *  it can often be prepared for a new compression with almost no work, only
+ *  sometimes falling back to the full, expensive reset that is always required
+ *  when the stream is in an indeterminate state (i.e., the reset performed by
+ *  LZ4_resetStreamHC()).
+ *
+ *  LZ4_streamHCs are guaranteed to be in a valid state when:
+ *  - returned from LZ4_createStreamHC()
+ *  - reset by LZ4_resetStreamHC()
+ *  - memset(stream, 0, sizeof(LZ4_streamHC_t))
+ *  - the stream was in a valid state and was reset by LZ4_resetStreamHC_fast()
+ *  - the stream was in a valid state and was then used in any compression call
+ *    that returned success
+ *  - the stream was in an indeterminate state and was used in a compression
+ *    call that fully reset the state (LZ4_compress_HC_extStateHC()) and that
+ *    returned success
+ *
+ *  Note:
+ *  A stream that was last used in a compression call that returned an error
+ *  may be passed to this function. However, it will be fully reset, which will
+ *  clear any existing history and settings from the context.
+ */
+LZ4LIB_STATIC_API void LZ4_resetStreamHC_fast(
+    LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel);
+
+/*! LZ4_compress_HC_extStateHC_fastReset() :
+ *  A variant of LZ4_compress_HC_extStateHC().
+ *
+ *  Using this variant avoids an expensive initialization step. It is only safe
+ *  to call if the state buffer is known to be correctly initialized already
+ *  (see above comment on LZ4_resetStreamHC_fast() for a definition of
+ *  "correctly initialized"). From a high level, the difference is that this
+ *  function initializes the provided state with a call to
+ *  LZ4_resetStreamHC_fast() while LZ4_compress_HC_extStateHC() starts with a
+ *  call to LZ4_resetStreamHC().
+ */
+LZ4LIB_STATIC_API int LZ4_compress_HC_extStateHC_fastReset (
+    void* state,
+    const char* src, char* dst,
+    int srcSize, int dstCapacity,
+    int compressionLevel);
+
+/*! LZ4_attach_HC_dictionary() :
+ *  This is an experimental API that allows for the efficient use of a
+ *  static dictionary many times.
+ *
+ *  Rather than re-loading the dictionary buffer into a working context before
+ *  each compression, or copying a pre-loaded dictionary's LZ4_streamHC_t into a
+ *  working LZ4_streamHC_t, this function introduces a no-copy setup mechanism,
+ *  in which the working stream references the dictionary stream in-place.
+ *
+ *  Several assumptions are made about the state of the dictionary stream.
+ *  Currently, only streams which have been prepared by LZ4_loadDictHC() should
+ *  be expected to work.
+ *
+ *  Alternatively, the provided dictionary stream pointer may be NULL, in which
+ *  case any existing dictionary stream is unset.
+ *
+ *  A dictionary should only be attached to a stream without any history (i.e.,
+ *  a stream that has just been reset).
+ *
+ *  The dictionary will remain attached to the working stream only for the
+ *  current stream session. Calls to LZ4_resetStreamHC(_fast) will remove the
+ *  dictionary context association from the working stream. The dictionary
+ *  stream (and source buffer) must remain in-place / accessible / unchanged
+ *  through the lifetime of the stream session.
+ */
+LZ4LIB_STATIC_API void LZ4_attach_HC_dictionary(
+          LZ4_streamHC_t *working_stream,
+    const LZ4_streamHC_t *dictionary_stream);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif   /* LZ4_HC_SLO_098092834 */
+#endif   /* LZ4_HC_STATIC_LINKING_ONLY */